Transcript LING001 - University of Pennsylvania

LING001
Evolution and Language
4-27-2009
1
Evolution and Language
•
•
“Nothing in biology makes sense except in the light of
evolution”.
Theodosius Dobzhansky
(1973)
The uniqueness of human language in the biological world
has driven people to “make sense” of it in the light of
evolution.
•
•
But it is important to distinguish mere “stories” from
testable scientific theories
Today:
•
•
basic structure of Natural Selection and the difficulties
with which it is applied to language and cognition
(Lewontin)
some suggestions about2the evolution of language
Angry Dick
•
•
Richard Lewontin (b. 1929): perhaps the greatest living
evolutionary biologist, also prominent social commentator
Outlines the structure of the argument for Natural
Selection and dismisses existing accounts of evolution of
language as story telling
3
Three Components of Natural
Selection
•
•
•
•
•
Variation: individuals must be different (no variation, no
evolution)
Heredity: differences must be passed down to
descendants
Selection: differential reproduction and survival due to
variation
The fallacy of the role of skin cancer in the evolution of
skin colors
Fitness is not constant: Bad eyesight is a good reason to
avoid the draft
4
Natural Selection in Real World
•
Natural Selection only increases the probability of passing
down the gene but doesn’t guarantee it: like a coin toss
•
•
•
•
N.S. can boost the chance of getting tails from 0.5 to
0.51 (0.01 advantage is considered large in nature)
If the population is very large, tails will make up close to
51% of the population in one generation, and more next,
and will gradually take over
But if the population is small, you may not have get more
tails in next generation even if it’s more likely to draw tails
Draw 10 balls: 60% of drawing red, 40% of drawing green
•
we may end up with a 6-4 split, but also 5-5, 2-8, even
0-10
5
Natural Selection in Real World
http://news.nationalgeographic.com/news/2008/04/080424-humans-extinct.html
•
For human evolution, it is generally held that while in
Africa, and when our ancestors marched out of Africa, the
population size was quite small
•
•
•
this makes the effect of sampling a lot more dramatic
as a result of this and other factors, in human
evolutionary genetics, the default hypothesis of
observing change is to assume that it is NOT caused by
Natural Selection
Which is why no serious biologists participate in the
story telling of the evolution of language and cognition:
it’s almost exclusively the game of psychologists,
philosophers, ... and linguists!
6
•
•
•
Stories
It’s possible to simply tell plausible stories about biological
properties of organisms.
In the case of language, there are many such stories;
language is required for communication, hunting, evolved
because of gossip
This ‘story-telling’ can
:
involve more or less
plausible parts; but
ultimately we’re interested in
theories that can be tested,
and that’s where most of
Lewontin’s critique is
directed.
No one doubts language did
evolve by biological mechanism
7
A “just so” story
•
•
Pinker & Bloom (1990: cited in Lewontin article)
Because language is complex (any questions?), it must be
the product of Natural Selection
•
•
Because Natural Selection has produced many complex
things though complexity is also the argument used by
creationists as evidence of a Designer
Much of evolutionary psychology has held Natural
Selection to be the force that created cognitive systems
•
•
“God gene”, “cheater detector gene”, etc., each with a
Natural Selection story that “makes sense”
But gene-trait mapping is not unique
8
Difficulties with Evolution of
• Language doesn’t Language
leave fossils
• We have no “close” relatives: chimps diverged from us
about 5-7 million years ago
•
•
The measurement of cognitive abilities is itself very
challenging
•
•
•
no time to evolve every component of language
individually, contrary to Pinker & Bloom
the Bell curve and the IQ test controversy
Not clear where to find genetically based variation
Not clear how variation translates into selectional
advantage
•
the first caveman who started talking might have been
killed instantly
9
Homology vs. Analogy
• Homology: similarity by common
descent
•
•
homology
•
Analogy: similarity via independent
evolutionary pathways
When a monkey is doing X, which
bears some rudimentary aspect of
language, it’s difficult to know X is
homology or analogy to infer about
the evolutionary history
With this in mind, we turn to a few
cases that have surfaced in language
and evolution in recent years that
compare the cognitive abilities of
humans
and other species
10
Speech Perception
•
Categorical perception: bah~pah
•
•
•
this has been disconfirmed: recall that chinchillas,
monkeys etc. can do CP as well
Tracking prosody to identify speech in language learning
•
•
once held to be uniquely human and uniquely linguistic
rodents and monkeys can do that too
But none of these touches on the core element of
language: the combinatorial infinity of phonology,
morphology, and syntax
11
Starling Syntax
(NY Times)
12
Monkey Business
Fitch & Hauser (2004, Science): cotton-top tamarin
A={ba, di, yo, tu, la, mi, no, wu}: female voice
B={pa, li, mo, ku, ka, bi, do, gu}: male voice
matched for duration, loudness
Are monkeys good at pattern matching?
13
Monkey Syntax?
Played a sequence of syllable, of the following patterns
n
(AB) = AB, ABAB, ABABAB ...
no li ba pa, la pa wu mo no li
n n
A B = AB, AABB, AAABBB
yo la pa do, ba la tu li pa ka
“Training” by exposure to a pattern
“Test” by playing a sequence than violates the training pattern
14
Monkey Syntax?
15
Monkey Syntax?
Such experiments are often difficult to interpret:
There could be other cognitive or processing mechanisms that affect
the ability to learn these patterns
But one possibility is that monkeys do not have language: they really
don’t
They lack the cognitive/computational capacity to process AnBn
patterns, which are the minimum requirement for the design of human
language
This may be the right kind of research question to pursue
16
Morphology
•
•
All languages have morphology
We create morphological classes from data with remarkable
accuracy
•
•
In languages like Italian or Spanish, children produce correct
agreement in over 98% of the contexts from the beginning of
speech
They seem to place words into their appropriate classes very
accurately
•
•
•
In English past tense, children over-regularize (about 10% of the
time) but they do not over-irregularize (about 0.2%)
i.e. they say “bring-bringed”, and do not say “bring-brang”
Categories are discrete rather than based overall similarity
(which would have produced think-thunk, flow-flew)
17
Categorization (Medin et al. 1987)
18
One dimensional sort
19
Learning in a maze
•
Many species learn to match the probabilities of
events/rewards/etc. in the environment when they make
choices
•
They do not, for instance, go for the most
prominent/frequent choice
20
Learning both languages
•
In “bilingual” environments, children acquire both options, closely
matching the frequencies of their usage
•
•
Here “bilingual” could mean two alternative choices in
phonology, morphology, or syntax, like some of the
sociolinguistic variation to be discussed on Wednesday
In monolingual environments, children go through a period of time
trying both the target and the non-target forms, which are given by
Universal Grammar, before settling on the target
•
•
These strongly suggest that at some level, the mechanisms of
language learning are the same as those used in running the
maze, and evolutionarily ancient
These mechanisms in turn pose constraints on, or explain, the
properties of human language
21
Conclusion
•
To understand language evolution, we need to partition the
properties of languages into distinct components and hope
to find them in other cognitive/perceptual systems and in
other species (Hauser, Chomsky & Fitch 2002)
•
•
The irreducible components must be uniquely human
and evolved recently (but don’t expect to trace them to a
single gene)
Of course, that begs the question of how the shared
components evolved, which may be cognitive and
leave little trace behind, and we are back to Lewontin.
22
•
•
Into the wild
There are many other linguistics courses that deal with
various subfields in depth: feel free to explore them.
The field of linguistics is becoming increasingly interdisciplinary
23